US20100332434A1 - Data classification tool using dynamic allocation of attribute weights - Google Patents
Data classification tool using dynamic allocation of attribute weights Download PDFInfo
- Publication number
- US20100332434A1 US20100332434A1 US12/459,367 US45936709A US2010332434A1 US 20100332434 A1 US20100332434 A1 US 20100332434A1 US 45936709 A US45936709 A US 45936709A US 2010332434 A1 US2010332434 A1 US 2010332434A1
- Authority
- US
- United States
- Prior art keywords
- attribute
- attributes
- conditional entropy
- variation
- interval
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 230000003068 static effect Effects 0.000 claims description 25
- 238000004364 calculation method Methods 0.000 claims description 6
- 230000009466 transformation Effects 0.000 claims description 4
- 230000001131 transforming effect Effects 0.000 claims description 4
- 238000013499 data model Methods 0.000 claims description 3
- 238000004458 analytical method Methods 0.000 abstract description 11
- 238000007635 classification algorithm Methods 0.000 abstract description 3
- 230000000694 effects Effects 0.000 abstract description 2
- 238000000034 method Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 238000004422 calculation algorithm Methods 0.000 description 3
- 238000013459 approach Methods 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 238000012706 support-vector machine Methods 0.000 description 2
- RBSXHDIPCIWOMG-UHFFFAOYSA-N 1-(4,6-dimethoxypyrimidin-2-yl)-3-(2-ethylsulfonylimidazo[1,2-a]pyridin-3-yl)sulfonylurea Chemical compound CCS(=O)(=O)C=1N=C2C=CC=CN2C=1S(=O)(=O)NC(=O)NC1=NC(OC)=CC(OC)=N1 RBSXHDIPCIWOMG-UHFFFAOYSA-N 0.000 description 1
- 230000004931 aggregating effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000000844 transformation Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/28—Databases characterised by their database models, e.g. relational or object models
- G06F16/284—Relational databases
- G06F16/285—Clustering or classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/08—Auctions
Definitions
- the present invention relates generally to an analysis tool and, more specifically, to a computer readable medium having instructions that cause a processor to analyze data using information gain of attributes and a classification algorithm to classify new records in a set of data.
- sorting incoming emails by content sorting data records as new data is entered, analyzing spend information and predicting an outcome based on a new item's characteristics.
- One such example is related to on-line auctions.
- an on-line auction be conducted electronically.
- Each participant in an auction does so through a remote site.
- the participant typically sends information from its remote site via a communications network (e.g. a network services provider) to an auction coordinator at its own remote site; the coordinator sends information to the participant in the same manner.
- the auction is actually managed by a software application which usually includes a client component operating at the remote site of the participant and a server component which typically operates at the coordinator's remote site.
- the client component communicates the bids entered by the participant to the server component.
- the server component communicates updates about the auction to the client component.
- the client component includes software used to connect to the server component, usually via telephone lines or the internet.
- the entity wishing to purchase services or items from suppliers may employ a service to manage auctions on its behalf or may employ special tools to assist with the management of its auctions.
- the service may employ tools that provide means to separate and select possible participants.
- the auction holder may be allowed to specifically designate certain suppliers it wishes the service to invite to the auction or to provide certain requirements for eligibility of suppliers which the service applies to determine the group of invitees.
- such services use tools that store information about past participants, both buyers and sellers, making the group of invitees simple to select based on the physical location of the past participants or on its inventory or offerings. Further, stored information can be used to assess the likely participation of a particular supplier and may even be applied to set a prebid ceiling or floor for that supplier based on its prior bidding behavior. It is not unusual to allow the entity wishing to purchase to set reserve prices, or to enter particular bid ceilings for a particular supplier with whom it has done business in the past or is the present incumbent.
- Another such example would include the need to sort and classify incoming emails. Accurate sorting can result in time savings as well as enhanced safety of a system. Being able to recognize spam accurately versus business or personal or enewsletters or desired emarketing would provide a level of efficiency related to email heretofore never provided.
- a record in a data set collected from a purchase, or an incoming email typically includes several attributes or categories of data.
- attributes in such data sets are often assigned a weight by the analyst according to a perceived level of importance the analyst believes that attribute may have in the data or that the analyst wishes to be able to use as a sorting factor in the set.
- the data itself may be considered in order to determine which attributes appear to have predictive value relative to other attributes. In either case, once an attribute is so analyzed it is assigned a weight and that weight is applied to incoming and new data.
- the classification tool of the present invention addresses the problems left unsolved by other systems.
- the present invention is an approach that includes 1) collecting comprehensive data from disparate sources 2) cleansing, reconciling, and aggregating the data 3) classifying the data into sourceable categories 4) and reporting the analysis and findings to the decision makers.
- the present invention determines the importance of each attribute by analyzing the pre-classified data. These are static weights of the attributes. Then, based on importance of each value within each attribute, it determines the interval of variation about the static weight of the attribute. Next, for the record to be classified, it determines the dynamic weight for each attribute depending upon the actual values of those attributes. Finally the present invention transforms the dynamic weights of the attributes into a prediction used to classify the new record.
- the present invention provides a way for its users to see accurate spend information at the aggregate and line item level for use in making time-sensitive decisions, perform spend analysis quickly and without dedication of additional resources, obtain visibility into how much money is spent with whom, for what, and how frequently, rationalize and standardize products and services, identify high performance suppliers and consolidate business to these high performance suppliers, monitor vendor compliance with contracted pricing and identify spend leakage, consolidate spend across various business units to increase leverage in supplier negotiations, identify and reduce maverick spend, identify product and service categories with the greatest savings potential, and understand key price trends and variances.
- Spend analysis by business unit, geography, category, vendor, price variance, etc. are available through the tool.
- the present invention includes instructions that, when read by a computer's processor, analyzes the pre-classified data in the system and applies a classification algorithm to transform that data and assign a new record to a particular classification.
- classification algorithm examples include posterior probability analysis, nearest neighborhood algorithms, and support vector machine.
- Posterior probability may be used relative to new data is a revised probability that takes into account new available information. For example, let there be two containers, container A having 5 black marbles and 10 red marbles and container B having 10 black balls and 5 red balls. If a container is selected at random, the probability that container A is chosen is 0.5. This is the a priori probability. If we are given an additional piece of information that a ball was drawn at random from the selected container, and that ball was black, then using posterior probability calculations we can determine the probability that the container from which it was drawn is container A. Posterior probability takes into account this additional information and revises the probability downward from 0.5 to 0.333 according to Bayes' theorem, because a black ball is more probable from container B than container A.
- the tool of the present invention first analyzes various values of attributes within a data set and then applies posterior probability to classify a new record.
- One way of looking at the data set is by analyzing and employing the analysis of an information gain of attribute.
- the information gain associated with an attribute is equal to the reduction in overall entropy of the system due to knowledge about that particular attribute and can be employed to assist in classifying a new record.
- the instructions of the present invention cause the computer to look at a particular pre-classified data set which includes records.
- Each record consists of a plurality of attributes or categories.
- a first attribute may have more predictive value than a second attribute but a third attribute may have more predictive value than the first attribute.
- the instructions cause the computer to assign to each of the attributes a static weight relative to their respective predictive values by calculating entropy of the data set, conditional entropy for each value of the attribute, transforming conditional entropy into information gain of the attribute and thereafter normalizing the information gain values.
- each attribute can be described by several different descriptors.
- Each descriptor of an attribute may have a different predictive value relative to a descriptor of another attribute or relative to another description of the same attribute.
- the invention instructs the computer to determine the interval of variation by first determining conditional entropy of each of the values for a particular attribute in the data set and transforming these values by inversion to obtain attribute importance. Next, these values are divided into quartiles and upper and lower boundary scaling factors are obtained which are then applied to find the upper bound and lower bound of the interval of variation of that attribute.
- the invention next provides a set of instructions to cause the computer to determine the dynamic weight of an attribute value in a particular record, specifically a record that is to be classified. For each attribute value in the new record, the conditional entropy in the data set is calculated for that value of the current attribute and the conditional entropy value is then transformed to obtain an Attribute value importance score. The attribute value importance score is then linearly transformed in the interval of variation for that attribute. Once the attribute value importance score for each attribute has been linearly transformed, they are normalized resulting in a dynamic weight measure for each attribute. Thereafter, the new record is assigned to a classification in light of the dynamic weight of each attribute value.
- the dynamic weight of each attribute is used as an exponent of corresponding likelihood probability value in Na ⁇ ve Bayes classification.
- the dynamic weight values could also be used in a Nearest Neighborhood algorithm which finds the nearest neighbor according to distance.
- the dynamic weight values would be used in a distance calculation.
- the dynamic weight values may be employed by a Support Vector Machine which operates on the premise of finding a hyper-plane with maximum width and then uses that hyper-plane for classification.
- the dynamic weight values would be used to determine the distance of the new record from the hyper-plane.
- FIG. 1 is a table providing a data set
- FIG. 2 is a flowchart of overall flow of the classification process
- FIG. 3 is a flowchart of process by which static weights are assigned to attributes
- FIG. 4 is a flowchart of process to find the interval of a variation of attributes of classification.
- FIG. 5 is a flowchart of process to assign dynamic weights to attributes of record to be classified.
- This computer readable medium 10 of the present invention comprises instructions 12 for a processor 13 that cause a computer 14 to make use of a pre-classified data set 16 (see FIG. 1 ).
- Data set 16 comprises a plurality of records 18 and that includes a plurality of attributes 20 - 21 wherein each of said plurality of attributes 20 - 21 for each record 18 may include one of a plurality of descriptors 32 - 33 and 24 - 26 for that attribute 20 - 21 .
- the computer 14 is also instructed to make use of a new record 40 to be classified. Generally, during classification, each attribute 20 - 21 is considered to be of equal importance.
- the importance of attributes 20 - 21 in deciding the classification may vary depending on distribution of the attribute descriptors 32 - 33 and 24 - 26 across the pre-classified data set and also depending on the attribute descriptors in the record 40 to be classified.
- An importance to an attribute 20 - 21 is typically translated into a weight assigned to that attribute.
- attributes play a role in accordance with their importance (or weights).
- Vendor Name 20 for this record set plays a more deciding role with certainty of the Category 22 by a larger amount than the knowledge of GL Description 21 .
- Vendor Name 20 attribute has more importance for deciding Category 22 , than GL Description 21 .
- Vendor Name 20 has value of “Company 1” 32 is different than the case where Vendor Name 20 has value of “Company 2” 33.
- Vendor Name 20 alone can decide Category 22 , whereas in later case Vendor Name 20 alone is not sufficient for the decision.
- Vendor Name 20 attribute is different in either of these cases.
- information gain 80 of an attribute 20 - 21 is equal to the reduction in overall entropy of the system due to knowledge of particular attribute 20 - 21 .
- the present invention takes into account the importance of attribute values 32 - 33 and 24 - 26 and varies the corresponding static weights 50 - 51 (or fixed importance), thus assigning the weights dynamically 62 - 63 to every attribute 20 - 21 of classification. Based on these dynamic weights 62 - 63 , a new record 40 can be classified using posterior probability calculations.
- the computer-readable medium 10 of the present invention contains the instructions 12 for a processor 13 that can be described generally as follows:
- the objective of the present invention is achieved by the computer readable medium 10 comprising the instructions 12 for a processor 13 to cause the computer 14 to perform the following transformations of attribute values and attribute descriptors to result in the predictive classification of the new record 40 .
- Static weights to Attributes Overall Attribute Conditional Information Normalized Entropy Name
- Entropy Gain Weights 0.4515 Vendor Name 0.1505 0.3010 1.5046 GL Description 0.3524 0.0991 0.4954 Get Interval for Variation about Static Weights 50 - 51
- Input new record 40 , Interval of Variation 60 - 61 for each attribute 20 - 21 in data set 16 , data set 16
- Dynamic Weights of Attributes Dynamic Attribute Attribute Attribute Weight Normalized Name Value Importance to Attribute Weights Vendor New Record 4 3.7 1.77 Name Vendor GL New Record 2.5 0.478 0.23 Description Supply
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Databases & Information Systems (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Strategic Management (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Economics (AREA)
- Development Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
Description
- The present invention relates generally to an analysis tool and, more specifically, to a computer readable medium having instructions that cause a processor to analyze data using information gain of attributes and a classification algorithm to classify new records in a set of data.
- There are a multitude of needs for means to classify items prospectively. For example, sorting incoming emails by content, sorting data records as new data is entered, analyzing spend information and predicting an outcome based on a new item's characteristics. One such example is related to on-line auctions.
- In the last ten years, on-line auction capabilities, features, and functions have evolved at lightning speed. Nearly as popular as forward auctions (auctions where bidders enter higher bids in order to win the item or service sold by the seller), reverse auctions (auction where bidders enter lower bids to entice a buyer to select the item or service sold by the bidder) have become the procurement tool of choice in many industries.
- It is typical that an on-line auction be conducted electronically. Each participant in an auction does so through a remote site. The participant typically sends information from its remote site via a communications network (e.g. a network services provider) to an auction coordinator at its own remote site; the coordinator sends information to the participant in the same manner. The auction is actually managed by a software application which usually includes a client component operating at the remote site of the participant and a server component which typically operates at the coordinator's remote site. The client component communicates the bids entered by the participant to the server component. The server component communicates updates about the auction to the client component. Most often the client component includes software used to connect to the server component, usually via telephone lines or the internet.
- The entity wishing to purchase services or items from suppliers may employ a service to manage auctions on its behalf or may employ special tools to assist with the management of its auctions. The service may employ tools that provide means to separate and select possible participants. For example, the auction holder may be allowed to specifically designate certain suppliers it wishes the service to invite to the auction or to provide certain requirements for eligibility of suppliers which the service applies to determine the group of invitees. Often, such services use tools that store information about past participants, both buyers and sellers, making the group of invitees simple to select based on the physical location of the past participants or on its inventory or offerings. Further, stored information can be used to assess the likely participation of a particular supplier and may even be applied to set a prebid ceiling or floor for that supplier based on its prior bidding behavior. It is not unusual to allow the entity wishing to purchase to set reserve prices, or to enter particular bid ceilings for a particular supplier with whom it has done business in the past or is the present incumbent.
- With all of the auctions being conducted, those hosting the auctions i.e. the buyers need to have tools that also assist them in analysis of the past and for prediction of the future relative to data previously collected and new data coming in. Further, all businesses whether they use electronic auctions to acquire supplies and services or not, have a need for specific analysis and reporting regarding their spend. Specifically, those entities that procure services and goods from many different suppliers need tools to assist them in analyzing what, how, and where they spend. Most companies have multiple enterprise transaction systems and lack the tools and processes to properly aggregate, cleanse, normalize and monitor their spend with multiple vendors at the line item level. This problem is particularly severe in the case of indirect goods and services, which typically bring a broader set of suppliers and considerably more transactions. Without having access to accurate, detailed spend data, organizations are unable to effectively source quality products and services at the best possible prices. What is needed is a solution that helps clients attain full visibility into their spend, better understand where savings opportunities exist, and how to best achieve the savings.
- Another such example would include the need to sort and classify incoming emails. Accurate sorting can result in time savings as well as enhanced safety of a system. Being able to recognize spam accurately versus business or personal or enewsletters or desired emarketing would provide a level of efficiency related to email heretofore never provided.
- A record in a data set collected from a purchase, or an incoming email, typically includes several attributes or categories of data. One can look at that data set in many ways in order to discern some pattern or to learn what that data means. However, attributes in such data sets are often assigned a weight by the analyst according to a perceived level of importance the analyst believes that attribute may have in the data or that the analyst wishes to be able to use as a sorting factor in the set. Alternatively, the data itself may be considered in order to determine which attributes appear to have predictive value relative to other attributes. In either case, once an attribute is so analyzed it is assigned a weight and that weight is applied to incoming and new data.
- While each of the aforementioned approaches has its value, neither takes into account the effect of a new record and, more particularly, on the weight or predictive value of any attribute. Therefore, left on the table is a significant level of informative data; the difference between a system that takes into consideration new record values relative to those already present as opposed to a static system.
- It is, therefore, one objective of the present invention to provide a tool for analyzing the predictive value of the attributes in records of a data set;
- It is another objective of the invention to provide a computer readable medium that, when employed by a computer's processor, analyzes the attributes and values of the attributes in a given set of data to obtain static weight or importance of each. Thereafter, the computer's processor varies those static weights or importance according to the values of those same attributes in a new record and readjusts that analysis as additional values are obtained from new records;
- It is a third objective of the invention to provide a tool that can be used to analyze data and provide reports regarding where money is spent across various aspects of a business, particularly at a line item level and in real-time.
- It is a fourth objective of the invention to provide a tool that can be used to analyze data and classify new data records into different classifications according to the new records attribute weights and dynamically adjusted attribute weights of the previously collected data.
- It is a fifth objective of the invention to provide a tool that can be used to analyze, classify and sort incoming emails to provide an efficient means of prioritizing emails for the recipient.
- The classification tool of the present invention addresses the problems left unsolved by other systems. The present invention is an approach that includes 1) collecting comprehensive data from disparate sources 2) cleansing, reconciling, and aggregating the data 3) classifying the data into sourceable categories 4) and reporting the analysis and findings to the decision makers.
- In its most general form, the present invention determines the importance of each attribute by analyzing the pre-classified data. These are static weights of the attributes. Then, based on importance of each value within each attribute, it determines the interval of variation about the static weight of the attribute. Next, for the record to be classified, it determines the dynamic weight for each attribute depending upon the actual values of those attributes. Finally the present invention transforms the dynamic weights of the attributes into a prediction used to classify the new record.
- The present invention provides a way for its users to see accurate spend information at the aggregate and line item level for use in making time-sensitive decisions, perform spend analysis quickly and without dedication of additional resources, obtain visibility into how much money is spent with whom, for what, and how frequently, rationalize and standardize products and services, identify high performance suppliers and consolidate business to these high performance suppliers, monitor vendor compliance with contracted pricing and identify spend leakage, consolidate spend across various business units to increase leverage in supplier negotiations, identify and reduce maverick spend, identify product and service categories with the greatest savings potential, and understand key price trends and variances. Spend analysis by business unit, geography, category, vendor, price variance, etc. are available through the tool.
- The present invention includes instructions that, when read by a computer's processor, analyzes the pre-classified data in the system and applies a classification algorithm to transform that data and assign a new record to a particular classification. Examples of such algorithms include posterior probability analysis, nearest neighborhood algorithms, and support vector machine.
- Posterior probability may be used relative to new data is a revised probability that takes into account new available information. For example, let there be two containers, container A having 5 black marbles and 10 red marbles and container B having 10 black balls and 5 red balls. If a container is selected at random, the probability that container A is chosen is 0.5. This is the a priori probability. If we are given an additional piece of information that a ball was drawn at random from the selected container, and that ball was black, then using posterior probability calculations we can determine the probability that the container from which it was drawn is container A. Posterior probability takes into account this additional information and revises the probability downward from 0.5 to 0.333 according to Bayes' theorem, because a black ball is more probable from container B than container A.
- The tool of the present invention first analyzes various values of attributes within a data set and then applies posterior probability to classify a new record. One way of looking at the data set is by analyzing and employing the analysis of an information gain of attribute. According to Information Theory, the information gain associated with an attribute is equal to the reduction in overall entropy of the system due to knowledge about that particular attribute and can be employed to assist in classifying a new record.
- At its most basic, the instructions of the present invention cause the computer to look at a particular pre-classified data set which includes records. Each record consists of a plurality of attributes or categories. A first attribute may have more predictive value than a second attribute but a third attribute may have more predictive value than the first attribute. The instructions cause the computer to assign to each of the attributes a static weight relative to their respective predictive values by calculating entropy of the data set, conditional entropy for each value of the attribute, transforming conditional entropy into information gain of the attribute and thereafter normalizing the information gain values.
- Further, each attribute can be described by several different descriptors. Each descriptor of an attribute may have a different predictive value relative to a descriptor of another attribute or relative to another description of the same attribute. Taking into account the variation of the descriptors around the attributes creates a dynamic weight for each attribute. Moreover, each time a new record is entered into the set having new values for each attribute, the dynamic weight of each attribute will change. Therefore, an interval of variation about the attribute caused by the values of that attribute must be considered and used in a determination of the dynamic weight of an attribute. The invention instructs the computer to determine the interval of variation by first determining conditional entropy of each of the values for a particular attribute in the data set and transforming these values by inversion to obtain attribute importance. Next, these values are divided into quartiles and upper and lower boundary scaling factors are obtained which are then applied to find the upper bound and lower bound of the interval of variation of that attribute.
- The invention next provides a set of instructions to cause the computer to determine the dynamic weight of an attribute value in a particular record, specifically a record that is to be classified. For each attribute value in the new record, the conditional entropy in the data set is calculated for that value of the current attribute and the conditional entropy value is then transformed to obtain an Attribute value importance score. The attribute value importance score is then linearly transformed in the interval of variation for that attribute. Once the attribute value importance score for each attribute has been linearly transformed, they are normalized resulting in a dynamic weight measure for each attribute. Thereafter, the new record is assigned to a classification in light of the dynamic weight of each attribute value. Employing posterior probability analysis, the dynamic weight of each attribute is used as an exponent of corresponding likelihood probability value in Naïve Bayes classification. The dynamic weight values could also be used in a Nearest Neighborhood algorithm which finds the nearest neighbor according to distance. The dynamic weight values would be used in a distance calculation. Alternatively, the dynamic weight values may be employed by a Support Vector Machine which operates on the premise of finding a hyper-plane with maximum width and then uses that hyper-plane for classification. Here, the dynamic weight values would be used to determine the distance of the new record from the hyper-plane.
- Other objects, features, and advantages of the present invention will be readily appreciated from the following description. The description makes reference to the accompanying drawings, which are provided for illustration of the preferred embodiment. However, such embodiment does not represent the full scope of the invention. The subject matter which the inventor does regard as his invention is particularly pointed out and distinctly claimed in the claims at the conclusion of this specification.
-
FIG. 1 is a table providing a data set; -
FIG. 2 is a flowchart of overall flow of the classification process; -
FIG. 3 is a flowchart of process by which static weights are assigned to attributes; -
FIG. 4 is a flowchart of process to find the interval of a variation of attributes of classification; and -
FIG. 5 is a flowchart of process to assign dynamic weights to attributes of record to be classified. - This computer readable medium 10 of the present invention comprises instructions 12 for a processor 13 that cause a computer 14 to make use of a pre-classified data set 16 (see
FIG. 1 ). Data set 16 comprises a plurality of records 18 and that includes a plurality of attributes 20-21 wherein each of said plurality of attributes 20-21 for each record 18 may include one of a plurality of descriptors 32-33 and 24-26 for that attribute 20-21. The computer 14 is also instructed to make use of a new record 40 to be classified. Generally, during classification, each attribute 20-21 is considered to be of equal importance. However, it is logical that the importance of attributes 20-21 in deciding the classification may vary depending on distribution of the attribute descriptors 32-33 and 24-26 across the pre-classified data set and also depending on the attribute descriptors in the record 40 to be classified. - An importance to an attribute 20-21 is typically translated into a weight assigned to that attribute. Traditionally equal weights have been assigned to attributes resulting in Static Weight Allocation; however, varying the weights (Dynamic Allocation of Weights) depending on the record to be classified, improves overall classification results. Thus for different records, attributes play a role in accordance with their importance (or weights).
- Example: As an example to illustrate this, consider the following said data set:
-
Vendor Name GL Description Category COMPANY 1 MRO NON-STOCK OTHER/MISC USAGE EQUIPMENT COMPANY 1 CAPITAL PROJECT OTHER/MISC EXP EQUIPMENT COMPANY 1 Spare Parts-SAP OTHER/ MISC EQUIPMENT COMPANY 1 MRO NON-STOCK OTHER/MISC USAGE EQUIPMENT COMPANY 2 Spare Parts-SAP CONVEYOR SUPPLIES COMPANY 2MRO NON-STOCK CONVEYOR SUPPLIES USAGE COMPANY 2 CAPITAL PROJECT POWER EXP TRANSMISSION COMPANY 2 Spare Parts-SAP POWER TRANSMISSION - Looking at the data, we can observe that knowledge of Vendor Name 20 for this record set plays a more deciding role with certainty of the Category 22 by a larger amount than the knowledge of GL Description 21. Thus we can claim that Vendor Name 20 attribute has more importance for deciding Category 22, than GL Description 21.
- Going further, the case where Vendor Name 20 has value of “
Company 1” 32 is different than the case where Vendor Name 20 has value of “Company 2” 33. In the earlier case, Vendor Name 20 alone can decide Category 22, whereas in later case Vendor Name 20 alone is not sufficient for the decision. Thus the importance of Vendor Name 20 attribute is different in either of these cases. - One mechanism to achieve the advantages of dynamic allocation of weight is by using information gain of attributes. In the Information Theoretic sense, information gain 80 of an attribute 20-21 is equal to the reduction in overall entropy of the system due to knowledge of particular attribute 20-21. The more the information gain 80, the more important is the attribute 20-21. This results in static weights 50-51 (or fixed importance) to the attributes 20-21. Then, for every new record 40 to classify, the present invention takes into account the importance of attribute values 32-33 and 24-26 and varies the corresponding static weights 50-51 (or fixed importance), thus assigning the weights dynamically 62-63 to every attribute 20-21 of classification. Based on these dynamic weights 62-63, a new record 40 can be classified using posterior probability calculations.
- The computer-readable medium 10 of the present invention contains the instructions 12 for a processor 13 that can be described generally as follows:
-
- 1. Read a pre-classified data set 16 including a plurality of attributes 20-21, along with its classification code 22 and, for each said attribute, a plurality of descriptors or values 32-33 and 24-26.
- 2. Calculate a Static Weight 50-51 for each of the plurality of attributes 20-21 in the data set 16.
- 3. Determine an Interval of Variation 60-61 for each of said plurality of attributes 20-21 in the data set 16.
- 4. For each new record 40 to be classified:
- a. Calculate a dynamic weight 62-63 for each of said plurality of attribute descriptors or values 32-33 and 24-26 using the static weight 50-51 and the interval of variation 60-61 for each of said plurality of attributes 20-21. This yields the attribute importance score of the current attribute by employing the actual attribute value in the new record. Then transform and normalize the set of attribute importance scores to obtain the dynamic weight of each of said plurality of attributes.
- b. Classify the new record 40 considering the dynamic weight 62-63 of each of said plurality of attribute values. For Naïve Bayes classification, the dynamic weight 62-63 value is used as an exponent of the individual conditional probabilities. The posterior probability based classification is thus as per:
-
-
-
-
- Wis are the Dynamic Weight 62-63 Values for each of the n attributes A1 to An.
-
-
- More specifically, the objective of the present invention is achieved by the computer readable medium 10 comprising the instructions 12 for a processor 13 to cause the computer 14 to perform the following transformations of attribute values and attribute descriptors to result in the predictive classification of the new record 40.
- Input: Pre-classified Data set 16 above
- Output: Static Weight 50-51 of each Attribute 20-21
- 1. Calculate overall entropy in the data set 16.
- 2. For each attribute 20-21 in the data set 16, do the following:
-
- a. For each distinct value of the attribute 32-33 and 24-26, calculate conditional entropy.
- b. Calculate weighted average of conditional entropy values and call it as the conditional entropy for the attribute 20-21.
- c. Calculate information gain 80-81 of the attribute 20-21 using the conditional entropy for the attribute 20-21 and the overall entropy.
Normalize the Information Gain 80-81 values of all the attributes 20-21, so that they sum up to the number of attributes 20-21. The normalized Information Gain 80-81 values now indicate the static weight 50-51 of each attribute 20-21. For the data set above, the calculations are:
-
Initial Entropy=−4/8*log(4/8)−2/8*log(2/8)−2/8*log(2/8)=0.4515 -
Entropy for Company 1(32)=−4/4*log(4/4)−0/4*log(0/4)−0/4*log(0/4)=0 -
Entropy for Company 2(33)=−2/4*log(2/4)−2/4*log(2/4)=0.3010 -
Total Entropy after knowledge of VendorName 20=0.5*Entropy(V1)+0.5*Entropy(V2) -
=0.5*0+0.5*0.3010=0.1505 -
Information Gain 80 for Vendor Name 20=Initial Entropy−Total Entropy after knowledge of VendorName -
=0.4515−0.1505=0.3010 -
Entropy for “MRO NON-STOCK USAGE”(30)=−2/3*log(2/3)−1/3*log(1/3)−0/3*log(0/3)=0.1174+0.159=0.2764 -
Entropy for “CAPITAL PROJECT EXP”(29)=−1/3*log(1/3)−1/3*log(1/3)−0/3*log(0/3)=0.159+0.159=0.318 -
Entropy for “Spare Parts-SAP”(28)=−1/4*log(1/4)−1/2*log(1/2)−1/2*log(1/2)=0.1505+0.1505+0.1505=0.4515 -
Total Entropy after knowledge of GL Description 21=3/8*0.2764+2/8*0.318+3/8*0.4515=0.3524 -
Information Gain 81 for GL Description 21=Initial Entropy−Total Entropy after knowledge of GL Description=0.4515−0.3524=0.0991 -
Un-normalized weights: Vendor Name 20=0.3010, GL Description 21=0.0991 -
Normalized weights: Vendor Name 20=1.5046, GL Description 21=0.4954 -
Static weights to Attributes Overall Attribute Conditional Information Normalized Entropy Name Entropy Gain Weights 0.4515 Vendor Name 0.1505 0.3010 1.5046 GL Description 0.3524 0.0991 0.4954
Get Interval for Variation about Static Weights 50-51 - Input: Data Set 16 above, Static Weights 50-51 of each attribute 20-21
- Output: Interval of Variation for each attribute 20-21 in the Data Set 16
- 1. For each attribute i 20-21 with static weight 50-51 in the Data Set 16:
-
- a. Find conditional entropy of each of the descriptors or values 32-33 and 24-26 for current attribute 20-21.
- b. (Note: Since conditional entropy indicates uncertainty about classification given some attribute value or descriptor, the lower the conditional entropy the lower would be uncertainty and hence more important would be the value or descriptor of attribute. Thus, the contribution of this value of attribute in the corresponding attribute's importance is inversely proportional to the conditional entropy.) One way to achieve this is to make Attribute value Importance score 85-86 as an inverse of the corresponding Conditional Entropy.
- c. Q1 is 25 percentile of Attribute value Importance scores 85-86, Q2 is 50 percentile of Attribute value Importance scores 85-86 and Q3 is 75 percentile of Attribute value Importance scores 85-86.
- d. Lower bound scaling factor for interval of variation 60 is calculated as: LBSFi=(Q2−Q1)/Q2.
- e. Lower bound of Interval of Variation 60 for current attribute 20-21 is calculated as: LBi=Wi−Wi·LBSFi
- f. Upper bound scaling factor for interval of variation 60 is calculated as: UBSFi=(Q3−Q2)/Q2.
- g. Upper bound of Interval of Variation 60 for current attribute 20-21 is calculated as: UBi=Wi+Wi*UBSFi.
For the data set provided above, the calculations are:
-
LBSF=(5−2)/5=0.6, LB=1.5046−1.5046*0.6=0.6018 -
UBSF=(10−5)/5=2, UB=1.5046+1.5046*2=4.5138 -
LBSF=(3−1.5)/3=0.5, LB=0.4954−0.4954*0.5=0.2477 -
UBSF=(5−3)/3=0.67, UB=0.4954+0.4954*0.67=0.8273 -
Interval of Variation Attribute Percentile Values Scaling Factor Bounds Name Q1 (25%) Q2 (50%) Q3 (75%) Lower Upper Lower Upper Vendor 2 5 10 0.6 2 0.6018 4.5138 Name GL 1.5 3 5 0.5 0.67 0.2477 0.8273 Description - Input: new record 40, Interval of Variation 60-61 for each attribute 20-21 in data set 16, data set 16
- Output: Dynamic Weights of each attribute 20-21:
- 1. For each attribute i 20-21 in the Record 40:
-
- a. Calculate conditional entropy in the pre-classified data set 16 for the descriptors or values of the current attribute 20-21. (Note: Since conditional entropy indicates uncertainty about classification given some attribute value, the lower the conditional entropy the lower would be uncertainty and hence more important would be the value of attribute. Thus, the contribution of this value of attribute in the corresponding attribute's importance score is inversely proportional to the conditional entropy.) One way to achieve this is to make Attribute Importance score 85-86 as an inverse of the corresponding Conditional Entropy.
- b. Attribute Importance score 85-86 is then transformed using Linear Transformation in the Interval of Variation 60-61 for current attribute 20-21. The transformed score is calculated using the formula:
-
-
-
- hi=Attribute Importance score 85-86
- Li=Lowest of Attribute value Importance scores 85-86 for Attribute i 20-21 in Data Model
- Hi=Highest of Attribute value Importance scores 85-86 for Attribute i 20-21 in Data Model
- UBi=Upper Bound of Interval of Variation 60-61 for Attribute i 20-21
- LBi=Lower Bound of Interval of Variation 60-61 for Attribute i 20-21
- th
i =Attribute importance score in Interval of Variation for Attribute i
- 2. Normalize the Attribute Importance scores 85-86 of all the attributes 20-21, so that they sum up to the number of attributes 20-21. The normalized values are the dynamic weights 62-63 of each of the attributes.
-
- Value=New Record Supply, Dynamic weight=0.478
-
-
Dynamic Weights of Attributes Dynamic Attribute Attribute Attribute Weight Normalized Name Value Importance to Attribute Weights Vendor New Record 4 3.7 1.77 Name Vendor GL New Record 2.5 0.478 0.23 Description Supply - Thus, the present invention has been described in an illustrative manner. It is to be understood that the terminology that has been used is intended to be in the nature of words of description rather than of limitation.
- Many modifications and variations of the present invention are possible in light of the above teachings. For example, any number of attributes and values of attributes may be considered. The instructions may partially be performed by several different computers or in several different stepwise configurations. Therefore, within the scope of the appended claims, the present invention may be practiced otherwise than as specifically described.
Claims (15)
LBSFi=(Q2−Q1)/Q2
UBSFi=(Q3−Q2)/Q2
LBi =W i −W i*LBSFi.
UBi =W i +W i*UBSFi.
LBSFi=(Q2−Q1)/Q2
UBSFi=(Q3−Q2)/Q2
LBi =W i −W i*LBSFi.
UBi =W i +W i*UBSFi.
Priority Applications (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US12/459,367 US8234230B2 (en) | 2009-06-30 | 2009-06-30 | Data classification tool using dynamic allocation of attribute weights |
US13/495,580 US9009091B2 (en) | 2009-06-30 | 2012-06-13 | Data classification tool using dynamic attribute weights and intervals of variation about static weights determined by conditional entropy of attribute descriptors |
US14/641,937 US20150317379A1 (en) | 2009-06-30 | 2015-03-09 | Data Classification Tool Using Dynamic Allocation of Attribute Weights |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US12/459,367 US8234230B2 (en) | 2009-06-30 | 2009-06-30 | Data classification tool using dynamic allocation of attribute weights |
Related Child Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US13/495,580 Continuation US9009091B2 (en) | 2009-06-30 | 2012-06-13 | Data classification tool using dynamic attribute weights and intervals of variation about static weights determined by conditional entropy of attribute descriptors |
Publications (2)
Publication Number | Publication Date |
---|---|
US20100332434A1 true US20100332434A1 (en) | 2010-12-30 |
US8234230B2 US8234230B2 (en) | 2012-07-31 |
Family
ID=43381817
Family Applications (3)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US12/459,367 Active 2031-01-17 US8234230B2 (en) | 2009-06-30 | 2009-06-30 | Data classification tool using dynamic allocation of attribute weights |
US13/495,580 Active 2030-06-03 US9009091B2 (en) | 2009-06-30 | 2012-06-13 | Data classification tool using dynamic attribute weights and intervals of variation about static weights determined by conditional entropy of attribute descriptors |
US14/641,937 Abandoned US20150317379A1 (en) | 2009-06-30 | 2015-03-09 | Data Classification Tool Using Dynamic Allocation of Attribute Weights |
Family Applications After (2)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US13/495,580 Active 2030-06-03 US9009091B2 (en) | 2009-06-30 | 2012-06-13 | Data classification tool using dynamic attribute weights and intervals of variation about static weights determined by conditional entropy of attribute descriptors |
US14/641,937 Abandoned US20150317379A1 (en) | 2009-06-30 | 2015-03-09 | Data Classification Tool Using Dynamic Allocation of Attribute Weights |
Country Status (1)
Country | Link |
---|---|
US (3) | US8234230B2 (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2013067079A1 (en) * | 2011-11-03 | 2013-05-10 | Microsoft Corporation | Systems and methods for handling attributes and intervals of big data |
WO2014193474A1 (en) * | 2013-05-31 | 2014-12-04 | Landmark Graphics Corporaton | Attribute importance determination |
US20150347950A1 (en) * | 2014-05-30 | 2015-12-03 | International Business Machines Corporation | Agent Ranking |
CN113807433A (en) * | 2021-09-16 | 2021-12-17 | 青岛中科曙光科技服务有限公司 | Data classification method and device, computer equipment and storage medium |
US11308098B2 (en) * | 2015-07-31 | 2022-04-19 | Priceline.Com Llc | Attribute translation and matching from a plurality of data records to create consolidated data records |
Families Citing this family (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20150221045A1 (en) * | 2014-01-31 | 2015-08-06 | Valify, LLC | System and method of normalizing vendor data |
US10127284B2 (en) * | 2014-12-18 | 2018-11-13 | International Business Machines Corporation | Scoring attributes in deep question answering systems based on algorithmic source code influences |
CN106384219B (en) * | 2016-10-13 | 2021-09-07 | 北京京东振世信息技术有限公司 | Storage sub-warehouse auxiliary analysis method and device |
US10523520B2 (en) | 2017-05-30 | 2019-12-31 | International Business Machines Corporation | System anomaly detection using parameter flows |
WO2020041335A1 (en) | 2018-08-21 | 2020-02-27 | Wal-Mart Stores, Inc. | Item line assignment systems and methods |
Citations (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20020013760A1 (en) * | 2000-03-31 | 2002-01-31 | Arti Arora | System and method for implementing electronic markets |
US20020013735A1 (en) * | 2000-03-31 | 2002-01-31 | Arti Arora | Electronic matching engine for matching desired characteristics with item attributes |
US20020042769A1 (en) * | 1999-03-31 | 2002-04-11 | Gujral Virind S. | System and method for conducting electronic auctions with multi-parameter optimal bidding |
US20020055900A1 (en) * | 2000-08-24 | 2002-05-09 | Namita Kansal | System and method of assessing and rating vendor risk and pricing of technology delivery insurance |
US20030014326A1 (en) * | 1999-06-23 | 2003-01-16 | Webango, Inc. | Method for buy-side bid management |
US20050091147A1 (en) * | 2003-10-23 | 2005-04-28 | Ingargiola Rosario M. | Intelligent agents for predictive modeling |
US20050192865A1 (en) * | 2004-02-24 | 2005-09-01 | Combinenet, Inc. | Automated scenario navigation in combinatorial exchanges |
US20050283425A1 (en) * | 2004-03-05 | 2005-12-22 | Russell Grove | Scalable auction management system with centralized commerce services |
US7003490B1 (en) * | 2000-07-19 | 2006-02-21 | Ge Capital Commercial Finance, Inc. | Multivariate responses using classification and regression trees systems and methods |
US20060069635A1 (en) * | 2002-09-12 | 2006-03-30 | Pranil Ram | Method of buying or selling items and a user interface to facilitate the same |
US20060167784A1 (en) * | 2004-09-10 | 2006-07-27 | Hoffberg Steven M | Game theoretic prioritization scheme for mobile ad hoc networks permitting hierarchal deference |
US20060282368A1 (en) * | 2005-06-10 | 2006-12-14 | Henry Cai | Adaptive stochastic transaction system |
US7212996B1 (en) * | 2000-04-20 | 2007-05-01 | Jpmorgan Chase Bank, N.A. | System and method for dynamic, multivariable comparison of financial products |
US20080052219A1 (en) * | 2006-03-31 | 2008-02-28 | Combinenet, Inc. | System for and method of expressive auctions of user events |
-
2009
- 2009-06-30 US US12/459,367 patent/US8234230B2/en active Active
-
2012
- 2012-06-13 US US13/495,580 patent/US9009091B2/en active Active
-
2015
- 2015-03-09 US US14/641,937 patent/US20150317379A1/en not_active Abandoned
Patent Citations (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20020042769A1 (en) * | 1999-03-31 | 2002-04-11 | Gujral Virind S. | System and method for conducting electronic auctions with multi-parameter optimal bidding |
US20030014326A1 (en) * | 1999-06-23 | 2003-01-16 | Webango, Inc. | Method for buy-side bid management |
US20020013735A1 (en) * | 2000-03-31 | 2002-01-31 | Arti Arora | Electronic matching engine for matching desired characteristics with item attributes |
US20020032638A1 (en) * | 2000-03-31 | 2002-03-14 | Arti Arora | Efficient interface for configuring an electronic market |
US20020013760A1 (en) * | 2000-03-31 | 2002-01-31 | Arti Arora | System and method for implementing electronic markets |
US7212996B1 (en) * | 2000-04-20 | 2007-05-01 | Jpmorgan Chase Bank, N.A. | System and method for dynamic, multivariable comparison of financial products |
US7433840B2 (en) * | 2000-04-20 | 2008-10-07 | Jpmorgan Chase Bank, N.A. | System and method for dynamic, multivariable comparison of financial products |
US7003490B1 (en) * | 2000-07-19 | 2006-02-21 | Ge Capital Commercial Finance, Inc. | Multivariate responses using classification and regression trees systems and methods |
US20020055900A1 (en) * | 2000-08-24 | 2002-05-09 | Namita Kansal | System and method of assessing and rating vendor risk and pricing of technology delivery insurance |
US6647374B2 (en) * | 2000-08-24 | 2003-11-11 | Namita Kansal | System and method of assessing and rating vendor risk and pricing of technology delivery insurance |
US20060069635A1 (en) * | 2002-09-12 | 2006-03-30 | Pranil Ram | Method of buying or selling items and a user interface to facilitate the same |
US20050091147A1 (en) * | 2003-10-23 | 2005-04-28 | Ingargiola Rosario M. | Intelligent agents for predictive modeling |
US20050192865A1 (en) * | 2004-02-24 | 2005-09-01 | Combinenet, Inc. | Automated scenario navigation in combinatorial exchanges |
US20050283425A1 (en) * | 2004-03-05 | 2005-12-22 | Russell Grove | Scalable auction management system with centralized commerce services |
US20060167784A1 (en) * | 2004-09-10 | 2006-07-27 | Hoffberg Steven M | Game theoretic prioritization scheme for mobile ad hoc networks permitting hierarchal deference |
US20060282368A1 (en) * | 2005-06-10 | 2006-12-14 | Henry Cai | Adaptive stochastic transaction system |
US20080052219A1 (en) * | 2006-03-31 | 2008-02-28 | Combinenet, Inc. | System for and method of expressive auctions of user events |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2013067079A1 (en) * | 2011-11-03 | 2013-05-10 | Microsoft Corporation | Systems and methods for handling attributes and intervals of big data |
WO2014193474A1 (en) * | 2013-05-31 | 2014-12-04 | Landmark Graphics Corporaton | Attribute importance determination |
CN105229494A (en) * | 2013-05-31 | 2016-01-06 | 界标制图有限公司 | Importance of Attributes is determined |
US9581726B2 (en) | 2013-05-31 | 2017-02-28 | Landmark Graphics Corporation | Attribute importance determination |
US20150347950A1 (en) * | 2014-05-30 | 2015-12-03 | International Business Machines Corporation | Agent Ranking |
US11308098B2 (en) * | 2015-07-31 | 2022-04-19 | Priceline.Com Llc | Attribute translation and matching from a plurality of data records to create consolidated data records |
CN113807433A (en) * | 2021-09-16 | 2021-12-17 | 青岛中科曙光科技服务有限公司 | Data classification method and device, computer equipment and storage medium |
Also Published As
Publication number | Publication date |
---|---|
US20120259867A1 (en) | 2012-10-11 |
US20150317379A1 (en) | 2015-11-05 |
US8234230B2 (en) | 2012-07-31 |
US9009091B2 (en) | 2015-04-14 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US8234230B2 (en) | Data classification tool using dynamic allocation of attribute weights | |
US11954623B2 (en) | Apparatus and method for resource allocation prediction and modeling, and resource acquisition offer generation, adjustment and approval | |
Deshpande et al. | Logistics performance, ratings, and its impact on customer purchasing behavior and sales in e-commerce platforms | |
US9009318B2 (en) | Offline resource allocation algorithms | |
Chen et al. | Auctioning keywords in online search | |
Chen-Ritzo et al. | Sales and operations planning in systems with order configuration uncertainty | |
US11216850B2 (en) | Predictive platform for determining incremental lift | |
Chen et al. | Assessing the competitiveness of insurance corporations using fuzzy correlation analysis and improved fuzzy modified TOPSIS | |
US20040177025A1 (en) | Real-time recommendations | |
US9836773B2 (en) | Evaluation and selection of quotes of a commerce network | |
KR20080086454A (en) | Data independent relevance evaluation utilizing cognitive concept relationship | |
US20180158004A1 (en) | Requesting Information from Organizations | |
CN109213936B (en) | Commodity searching method and device | |
CN112598472A (en) | Product recommendation method, device, system, medium and program product | |
WO2022257731A1 (en) | Method, device and system for performing algorithm negotiation on privacy computation | |
US10346868B2 (en) | Gift exchange platform | |
US20170039196A1 (en) | Computing system to evaluate sentiment of entities by processing data | |
Kachani et al. | Competitive Pricing in a Multi‐Product Multi‐Attribute Environment | |
US20120265588A1 (en) | System and method for recommending new connections in an advertising exchange | |
US8688537B2 (en) | Maintenance of a company profile of a company associated with a supplier/buyer commerce network | |
Scott et al. | Lemon complaints: when online auctions go sour | |
Ekström et al. | Accounting for rater credibility when evaluating AEC subcontractors | |
Nadanyiova et al. | Financing of adverts and its impact on the brand value | |
CN114493132A (en) | Resource allocation method and device and electronic equipment | |
Li et al. | A bid evaluation method for multi-attribute online reverse auction |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
AS | Assignment |
Owner name: GLOBAL EPROCURE, NEW JERSEY Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:PAWAR, SACHIN SHARAD;JOSHI, GIRISH;REEL/FRAME:022936/0844 Effective date: 20090527 |
|
STCF | Information on status: patent grant |
Free format text: PATENTED CASE |
|
AS | Assignment |
Owner name: CITIBANK, N.A., CALIFORNIA Free format text: SECURITY INTEREST;ASSIGNORS:NB VENTURES, INC.;GEP (GLOBAL EPROCURE);REEL/FRAME:034007/0585 Effective date: 20141017 |
|
FPAY | Fee payment |
Year of fee payment: 4 |
|
FEPP | Fee payment procedure |
Free format text: ENTITY STATUS SET TO UNDISCOUNTED (ORIGINAL EVENT CODE: BIG.); ENTITY STATUS OF PATENT OWNER: LARGE ENTITY |
|
MAFP | Maintenance fee payment |
Free format text: PAYMENT OF MAINTENANCE FEE, 8TH YEAR, LARGE ENTITY (ORIGINAL EVENT CODE: M1552); ENTITY STATUS OF PATENT OWNER: LARGE ENTITY Year of fee payment: 8 |
|
MAFP | Maintenance fee payment |
Free format text: PAYMENT OF MAINTENANCE FEE, 12TH YEAR, LARGE ENTITY (ORIGINAL EVENT CODE: M1553); ENTITY STATUS OF PATENT OWNER: LARGE ENTITY Year of fee payment: 12 |